Stock Analysis Platform
A cost-safe multi-agent stock research demo that turns deterministic market snapshots into chart-backed recommendations.
5 agents
Analysis model
Fundamental, technical, sentiment, risk, and market-context perspectives are presented as separate specialist signals before the final recommendation.
8 tickers
Demo coverage
The public demo supports allowlisted snapshots for AAPL, MSFT, NVDA, TSLA, AMZN, GOOGL, META, and NFLX.
No keys
Public-safe architecture
Visitors do not enter API keys, and the deployed demo does not call paid LLM, stock, news, or database-write services.
Problem
AI finance demos can be expensive, flaky, or risky when they depend on live APIs, visitor-provided keys, or third-party market services during a portfolio review.
Approach
Built a public demo mode around allowlisted stock snapshots, a small API surface, five specialist analysis perspectives, chart context, and clear unsupported-symbol handling.
Impact
Recruiters and reviewers can explore the full product experience quickly without API keys, surprise costs, database writes, or live-service failures.
Overview
This project focuses on a practical portfolio question: how do you make an AI-assisted stock analysis experience useful, reliable, and safe for public review?
Instead of asking visitors for API keys or depending on live paid services, the public version uses pre-generated stock snapshots. That keeps the demo fast and deterministic while still showing the product workflow, agent model, charting, consensus logic, and recommendation presentation.
Product shape
The homepage centers on a supported ticker flow. Once a symbol is submitted, the app loads a structured demo snapshot and renders:
- stock overview
- interactive price chart
- market context notes
- five-agent analysis results
- consensus score and final recommendation
That structure matters because the recommendation is anchored to visible evidence instead of appearing as a single opaque answer.
Why the architecture is interesting
The project is more than a UI shell around an answer box. It treats the public demo itself as a reliability problem.
Multi-agent reasoning
The platform breaks the problem into distinct perspectives:
- fundamental signals
- technical signals
- sentiment
- risk
- market context
That separation makes the result easier to reason about than a single monolithic response.
Cost-safe public mode
The hosted version is designed for reviewers:
- no visitor API keys
- no paid AI calls
- no stock or news API dependency
- no database mutation routes
- clear allowlist for supported symbols
That makes the project stronger as a portfolio piece because the user experience is predictable when someone opens it from a resume link.
Product workflow
The page design supports a clear reviewer sequence:
1. search for a symbol 2. review core stock data 3. inspect price movement 4. scan market context 5. read agent reasoning, consensus, recommendation, confidence, and risk
Quality and reliability considerations
Financial demos need clearer boundaries than a typical toy AI app.
Important quality considerations here include:
- allowlisting supported tickers
- returning useful errors for unsupported symbols
- keeping demo data deterministic
- avoiding exposed secrets
- making clear that the content is demo-only and not financial advice
Why it belongs in the portfolio
This is a strong public-facing project because it shows product judgment, orchestration thinking, frontend execution, and deployment discipline in one place. It demonstrates an AI workflow without making the reviewer depend on live AI infrastructure.
Project proof
Stack